There are a number of elegant studies that have been developed over the past two decades investigating links between the motor cortex and language, and although they are reviewed more comprehensively elsewhere in this thesis, below is a brief overview. Results from neuroimaging studies demonstrate increased activity in the areas of the motor cortex when processing action-related language. Results from fMRI studies indicate overlapping activation patterns in the effector-specific regions in the motor cortex when reading action verbs and performing the described action (Hauk et al., 2004) or when reading phrases that included a specific effector and observing that action (Aziz-Zadeh et al., 2006). Even without the comparison of action observation, neuroimaging researchers have reported increased motor cortex activity when listening to sentences that describe actions compared to abstract sentence (Tettamanti et al., 2005), or when participants perceived action words (nouns or verbs) compared to non-action words (Vigliocco et al., 2006).
The notion that the motor cortex in involved in processing motor-related language has led some researchers to investigate how motor cortex activity could be measured using TMS – a technique used widely in motor control research in which researchers induce EMG activity data and interpret it as reflection of motor cortex cortical activity. Buccino et al. (2005) used TMS to compare MEP (a specific component of EMG) amplitudes when listening to either hand-based or foot-based action sentences, or abstract sentences. Buccino et al. found that when listening to a hand-based action sentence, MEPs were significantly lower in the hand muscles compared to foot-based action and abstract sentences. Similarly, MEPs were
significantly lower in the foot muscles when listening to foot-based action sentences compared with the other two conditions. Based on the notion that larger MEPs are indicative of higher excitability of the corticospinal pathway and smaller MEPs indicate lower excitability (Rothwell, 1997), the researchers interpret these excitability fluctuations as evidence that the motor cortex is involved in action- language processing. However, they acknowledge that it would be expected that there would have been an increase (rather than the observed decrease) in MEP amplitude as demonstrated in other action observation studies (see Buccino et al., 2005, for a review).
Oliveri et al. (2004a) conducted a series of TMS studies investigating the role of motor cortical regions on the processing of action-related words. In one of these studies, participants were required to produce utterances comprising of action verbs (e.g., 'to strangle'), action nouns (e.g., 'the axe'), non-action verbs (e.g., 'to adore'), and non-action nouns (e.g., 'the carpet'). Using paired-pulse TMS, Oliveri et al. assessed changes in motor cortex excitability when participants produced a list of utterances including these different word types. Paired-pulse methodology generally consists of administration of a subthreshold conditioning stimulus followed by a suprathreshold test stimulus. Based on the time between pulses and variations in MEP size, researchers are able to infer changes in the excitability of intracortical inhibitory and facilitatory circuits (Oliveri et al., 2004a). Measuring MEP size, the results showed that when TMS was applied over the hand-area of the motor cortex, there was a significant increase in excitability when producing motor related words compared to non-motor words. Interestingly, Oliveri et al. report that this effect transcends grammatical class division as increased excitability was observed in both action-noun and action-verbs, and there was no significant difference between word
classes within either action-type. The authors cite this as evidence that the motor cortex is stimulated by more than just description of an executable action, and that processing words that merely describe an object that would be involved in an action also recruits the motor cortex. While Oliveri et al. do not draw a direct connection to mirror neuron activity, these findings fit within a broader mirror neuron framework and are concurrent with embodied cognition theories (Gallese & Lakoff, 2005) and the links between the motor cortex and imagining object-specific actions.
While Oliveri et al. (2004a) state that they wouldn't necessarily expect similar action versus non-action results using single-pulse TMS methodology, other researchers have found that single-pulse stimulation can also produce facilitative differences in reaction time (RT) to action words, depending on the region of the cortex stimulated, and the effector being used in the described action. As covered in the TMS chapter of this thesis (Chapter 5), single-pulse TMS can be applied during a task with the intention of impacting stimulus processing and change behavioural outcome. Pulvermüller et al. (2005) set out to investigate the impact of applying single-pulse TMS to the motor cortex of participants while they viewed string of words consisting of arm actions, leg actions, distracter words and pseudowords. A TMS pulse was delivered at 90% of resting motor threshold (RMT) over the left motor cortex 150 ms after the presentation of each item while EMG data were recorded from a muscle in the contralateral forearm as well as the leg. The choice of delivering the TMS pulse at 150 ms in this study is based on previous reports that the time period during which semantic processing is thought to start is 100 to 200 ms (Pulvermüller, 2001; Skrandies, 1998). The pulse was delivered at an intensity of 90% of the RMT with the expectation that this would prime motor cortex
Participants were asked to respond to meaningful words by making a brisk mouth movement and accuracy and RT was calculated for each type of word. Pulvermüller et al. observed faster RT to arm action words compared to leg action words after TMS of the area of motor cortex responsible for the arm, and vice versa for TMS of the area of the motor cortex responsible for the leg. This finding was interpreted as evidence that words that are semantically linked to hand or leg actions are processed by the effector-specific area of the motor cortex responsible for controlling those limbs.
Based on an adaptation of the Pulvermüller et al. (2005) methodology, in a previous study (Mertens, 2009), I set out to investigate the effect of a supra-threshold TMS pulse on the processing of hand-action words in the motor cortex. In that study, participants viewed a list of hand-action and non-action (or ‘static’) verbs while single pulse TMS was applied to the hand area of the motor cortex. Based on the findings that a sub-threshold TMS pulse applied at 150 ms facilitates the processing of motor-action verbs (Pulvermüller et al., 2005), and previous reports of supra- threshold TMS delaying reaction time (Pascual-Leone et al., 1992) I wished to determine if a supra-threshold TMS pulse of 120% of RMT applied to the hand area of the motor cortex would inhibit processing of hand-action verbs, and therefore delay reaction time (RT). In that study, 15 participants viewed a list of words that consisted of a mix of hand-action verbs (such as wipe, or carry), staticverbs (such as
dream, or fear), and pseudowords which were used as distracters. Participants were required to produce a quick jaw-clench response only upon viewing a real English word, and differences in RT between TMS and sham (control) trials were analysed. Although RT was significantly slower for action-verbs, application of TMS yielded no significant change to RT for either type of word. It could therefore be inferred
that as a TMS pulse delivered to the hand-area of the motor cortex at 120% of RMT at 150 ms post stimulus onset does not create any discernible delay in RT, this protocol does not cause inhibition in the processing of hand-action words. Based on these results, I was unable to conclusively rule out the notion that supra-threshold TMS causing inhibition to motor-cortex processing of motor-related words, but instead concluded that the lack of findings in this study may be due to the pulse latency not targeting the appropriate stage of language processing to produce inhibitory effects. Furthermore, I also suggested that the facilitatory effect found by Pulvermüller et al. was specific to the application of a pulse of 90% of RMT
delivered at 150 ms post stimulus onset.
The current study aims to further elaborate on the findings of my earlier study (Mertens, 2009) by investigating the effects of single pulse TMS applied over the hand-area of the motor cortex at a range of different intensities and latencies. To determine an appropriate range of pulse latencies, it is necessary to further explore the literature examining the time course of language processing throughout linguistic cortical networks. While not investigating neural activity in response to motor- related words, Skrandies (1998) examined ERP component divergence due to semantic differentiation. In that study, EEG data were recorded from participants who viewed 60 German nouns that had been previously rated as 'high' on one aspect of one of three dimensions; evaluation (good-bad); potency (strong-weak); and
activity (active-passive). In order to ensure participants were attending, they were asked to actively remember and visualise the words presented as their memory would later tested. Skrandies specifically investigated the P100 ERP component - a positive component occurring around 100 ms post stimulus onset. It was observed that compared to control (viewing a checkboard pattern), word stimuli elicited a
delayed P100 peak latency (by approximately 4 ms). Furthermore, words that ranked high or low on the activity dimension elicited later P100 peaks, compared to words ranked moderately on this scale. Skrandies also observed changes in other sections of the ERP waveform and found that words high or low in the potency dimension elicited greater amplitudes of a negative component at approximately 160 ms, and a delayed peak latency for a negative component occurring at approximately 220 ms. While Skrandies also reports some further distinctions in later components, he concluded that most of the significant early semantic effects were observed at around the latency of 100 ms. Skrandies admits this latency is early compared to other studies which report semantic processing occurring much later (such as the N400 reported by Kutas and Hillyard [1980]). While he cautions against interpreting reading rate studies alongside his electrophysiological results, Skrandies explains that early semantic processing is quite possible as there is evidence that the human brain can rapidly process words when speed reading (as Rubin and Turano [1992] report that some participants can read at up to 30 words per second).
As indicated by Skrandies (1992) and outlined earlier in this thesis, EEG studies have previously identified the much later N400 component as reflective of
processing linguistic information (for a review see Kutas & Federmeier, 2011). In many studies the N400 peak appears to be attenuated by attempts to integrate
semantic information from a linguistic stimulus. Modulations in N400 amplitude can be driven by a participant's expectations of a stimulus, or more accurately, the
incongruency of the semantic link between a priming stimulus and the subsequent target stimulus in both word pairs and sentences (Kutas, 1993). In a comprehensive review of N400 literature, Kutas and Federmeier are cautious to point out that
linked to pure linguistic processing and modulations in N400 can be seen in studies that use a broad range of less linguistically salient stimuli (such as pictures, or actions) and can also be used to examine aspects of semantic memory (Kutas & Federmeier, 2011).
In a study more relevant to the time-course of processing motor-related words, Kellenbach, Wijers, Hovius, Mulder, and Mulder (2002) examined
differences in ERP components following the presentation of words from different grammatical classes and different semantic categories. In this study, Kellenbach et al. asked Dutch participants to attentively read nouns and verbs that belonged to one of three semantic categories: abstract (e.g., 'klemtoon', emphasis, and 'bedroeven', grieve); visual (e.g., 'nummerbord', license plate, and 'dwarrelen', whirl); and motor (e.g., 'pincet', tweezers and 'afwassen', wash). Kellenbach et al. analysed 50 ms sections of ERP waveforms that occurred following the presentations of these words looking for component differences elicited by different word types. The researchers report no interaction effects between grammatical class and word type, though they did observe that the earliest divergence for motor related words compared to abstract and visual words occurred at the 250 ms to 450 ms range. Although their ERP data did not indicate a specific cortical region involved in this process, Kellenbach et al. suggest that this is indicative that processing of semantic content of motor words occurs around this time-point. They go on to say that these data are evidence that "motor knowledge is accessed particularly quickly or efficiently, regardless of whether the stimulus is a manipulable object (noun) or an action (verb)" (p. 571). Kellenbach et al. propose that this rapid processing of motor-related words is due to the high salience of motor-word attributes (such as the cross-modal imageability of the object and associated action) compared to uni-modal visual words or abstract
words. Regardless of whether a word’s 'motor salience' is truly the reason that motor-words are processed at this time point, this research still provides evidence suggesting that the 250 ms to 450 ms epoch likely encapsulates the earliest processing of content specific to motor words.
Like Kellenbach et al. (2002), other researchers have sought to investigate the time-course of processing motor related language and as per Pulvermüller et al. (2005), TMS has been used as a reliable method of altering motor cortex function in order to examine its role in action-language processing. Papeo, Vallesi, Isaja, and Rumiati (2009) examined the effect of TMS pulses delivered over the motor cortex at 120% of RMT at 170, 350, and 500 ms post stimulus onset. Papeo et al. examined changes in MEP amplitude as well as response accuracy and RT in trials where participants were either required to explicitly identify whether verbs were action- related or not (semantic task), or were required to decide on the number of syllables in the verbs (syllabic task). The verbs presented (in Italian) to the participants described either hand-actions (e.g., ‘mescolo’, I stir), non-hand actions (e.g., ‘salto’,
I jump), or non-action (e.g., ‘medito’, I wonder). At all three latencies, accuracy was not significantly different among word types or between sham and TMS.
Additionally, although no difference in RT was observed when the TMS pulse was delivered at 170 ms, a pulse at 350 ms led to slower RT compared to sham (control) to non-action words in the semantic task, and to both hand-action and non-hand action words in the syllabic task. When examining MEP results, Papeo et al. report that a TMS pulse delivered at 500 ms post-stimulus onset leads to increased MEP amplitude for hand action words when performing the semantic task compared to the syllabic task. As this effect is not seen elsewhere (at other latencies or for other word types), Papeo et al. suggest that the increased MEP amplitude at this time point in the
semantic task represents a facilitation in motor cortex activity when processing semantic information about hand-action words. Unfortunately, as the semantic and syllabic tasks are inherently different, RT data cannot be compared in order to detect if this increased facilitation impacts behavioural responses. That said, there was also no significant difference between RT in sham and TMS conditions for both tasks when the pulse was delivered at 500 ms, which indicates that this facilitation of the motor cortex during the semantic task is not linked to any significant TMS-induced RT change. Despite increased MEP amplitude at 500 ms, it is difficult to confidently conclude that this would be the ideal latency at which TMS should be applied to induce facilitation or inhibition, particularly without supporting behavioural results. Despite mixed results, as there were differences in the way TMS impacted the semantic versus the syllabic task, Papeo et al. were able to conclude that although there was evidence that the motor cortex is involved in processing semantic content of effector-specific action-related verbs. The author suggest that this process is not automatic upon perception of the word, and instead relies on some deliberate processing of semantic decision-making information in order to recruit motor networks.
Beyond lexical decision-making tasks, to gain a greater insight into the impacts of various single-pulse TMS parameters on behavioural response, it is also important to review the impact of TMS on simple RT. Pascual-Leone et al. (1992) conducted a study that examined the effects of single pulse TMS over the motor cortex with varied latency and intensity on RT in a simple go-signal task.
Participants were asked to flex their right thumb or elbow as quickly as possible in response to a 'go' signal that was presented in either a visual, auditory, or tactile modality. RT was measured as the onset of EMG activity recorded from either the
bicep or abductor pollicis brevis (APB) muscle in the thumb. Pascual-Leone et al. administered a sub-threshold TMS pulse concurrently with the go-signal as well as intervals of 5 or 10 ms either side of the go-signal up to ±50ms. In that study, 'sub- threshold' intensity was below resting motor threshold (RMT) and was defined as the greatest intensity at which no MEPs were elicited that were larger than 50µV peak- to-peak. All results were compared against control condition, in which participants responded to the go signal without concurrent application of TMS. It was found that fastest reaction times could be seen at +5 to +10 ms post-stimulus onset, and
compared to control trials, TMS at this latency improved RT by approximately 30 ms. The facilitative effect of TMS on RT is less apparent at latencies of -10 to -50 ms and +20 to +40 ms and was observed to actually inhibit/delay RT when TMS is delivered at +50ms. Pascual-Leone et al. also examined the impact of altering TMS intensity on RT by administering it at intervals of ±5% of sub-threshold intensity. RT was fastest (approx 25 ms faster than control) at sub-threshold intensity as well as 5% below. Again, this facilitative effect was much less discernible at -15% to -20% and +5% to +10% of sub-threshold intensity, and actually delayed RT at +15% to +25%.
It is difficult to directly compare the findings of Pascual-Leone et al. (1992) with Pulvermüller et al. (2005), as they used different methods to determine their TMS pulse intensity. As noted above, Pascual-Leone et al. tested using a sub- threshold intensity defined as the maximum intensity at which no MEPs were elicited, whereas Pulvermüller et al. found the minimum intensity at which MEPs were elicited (RMT) and then tested at 90% of that value. Despite the differences in methodology, it can comfortably be acknowledged that in both studies sub-threshold intensities promoted faster RT. Interestingly facilitation was seen in both the simple
go-signal task (Pascual-Leone et al., 1992) as well as in the vastly more complex linguistic decision making task (Pulvermüller et al., 2005). While it may be tempting to interpret this as an indication that a sub-threshold pulse to the motor cortex
facilitates action regardless of complexity or modality of the task being attempted, it